Aiming at segmentation to images with nonuniform gray-level, an image segmentation method adaptive to gray value was proposed.Based on the level set theory, by using the gray value information of target image and the background, adaptive linear fitting was made to the global and local information.Then, image segmentation was conducted to avoid being trapped in the local optimum, which was robust to noise.The experimental results showed that the proposed method can resist noise effectively, is adaptive to the nonuniform gray level information, and can obtain global optimal image segmentation.